Add official and reproducible benchmarks for the CSV readers

- [ ] choose right tooling for CSV reader benchmarking
- [ ] prepare a few workload scenarious
- [ ] compare our CSV readers between each other
- [ ] compare with the other (not only JVM readers) if it's possible
- [ ] posting results on the GitHub
- [ ] refer to that in our official documentation
- [ ] share the code and make it reproducible for the community